Zhang Et Al-2002-Earthquake Engineering & Structural Dynamics

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2002; 31:19671991 (DOI: 10.1002/eqe.197)
Seismic response analysis of highway overcrossings including

soilstructure interaction
Jian Zhang and Nicos Makris;

Department of Civil and Environmental Engineering; University of California; Berkeley; CA 94720-1710; U.S.A.
SUMMARY
This paper presents a systematic procedure for the seismic response analysis of highway overcrossings.
The study employs an elementary stick model and a more sophisticated nite element formulation to
compute response quantities. All dynamic stinesses of approach embankments and pile groups are
approximated with frequency-independent springs and dashpots that have been established elsewhere. A
real eigenvalue analysis conrms the one-to-one correspondence between modal characteristics obtained
with the three-dimensional nite element solutions and the result of the simpler stick-model idealization.
A complex eigenvalue analysis yields modal damping values in the rst six modes of interest and shows
that modal damping ratios assume values much higher than those used by Caltrans. The validity of
the proposed method is examined by comparing the computed time response quantities with records
from the Meloland Road and Painter Street overcrossings located in southern and northern California,
respectively. The proposed procedure allows for inexpensive parametric analysis that examines the
importance of considering soilstructure interaction at the end abutments and centre bent. Results and
recommendations presented by past investigations are revisited and integrated in comprehensive tables
that improve our understanding of the dynamic characteristics and behaviour of freeway overcrossings.
The study concludes with a step-by-step methodology that allows for a simple, yet dependable dynamic
analysis of freeway overcrossings, that involves a stick model and frequency-independent springs and
dashpots. Copyright ? 2002 John Wiley & Sons, Ltd.
KEY WORDS: dynamic analysis, highway bridges, earthquakes, soilstructure interaction
INTRODUCTION
The motivation for this study did not originate from a lack of published material on the
seismic response of freeway overcrossings but rather from an abundance of publications,
the vast numbers of which present ndings that are scattered, occasionally conicting, and
Correspondence to: Nicos Makris, Department of Civil and Environmental Engineering, University of California,
Berkeley, CA 94720-1710, U.S.A.
E-mail: makris@ce.berkeley.edu
Contract=grant sponsor: National Science Foundation; contract=grant number: CMS 9696241.
Received 9 May 2001

Revised 3 October 2001 and 6 February 2002
Copyright ? 2002 John Wiley & Sons, Ltd. Accepted 6 February 2002
1968 J. ZHANG AND N. MAKRIS
derived from various methodologies that in many occasions have little in common. As a
result, despite several existing recommendations (References [1; 2], among others) there is
no established procedure that results in a dependable estimate of the seismic response of
freeway overcrossings, partly because the ndings of the below-mentioned studies have not
been combined in a rational manner that will result in a systematic analysis procedure.
The seismic response of freeway overcrossings received distinct attention in the late 1980s.
Maragakis and Jennings [3] introduced the stick model enhanced with bilinear springs and
dashpots at its support to study the motion of skew overpasses. Their model accounted for
several practical diculties such as the presence of elastomeric pads and the gap between
the deck and the back wall. Werner et al. [4] developed a system identication methodology
to extract information from an array of strong-motion measurements that were recorded in
the vicinity of the Meloland Road Overcrossing during the 1979 Imperial Valley earthquake.
Their conclusions emphasized the ability of linear models to t the measured response and
the pronounced eects that the approach embankments and foundations have on the response
of the bridge. Although their paper identies relatively low values of modal damping for
the bridgefoundation system (i =68%), a later publication by Werner [1] indicates modal
damping ratios ranging from 19 to 26%. About the same time, Crouse et al. [5] conducted ex-
perimental and analytical studies to determine the signicance of soilstructure interaction on
the response of a single span overcrossing with monolithic abutments on spread footings. The
small displacement gradient generated from the ambient quick-release and forced-vibration
tests resulted in small values of damping and large values of stiness that are not represen-
tative under earthquake loading. The present paper revisits the problem of seismic response
of highway overcrossings and proposes a systematic procedure to account for soilstructure
interaction. Our goal is not to derive another isolated study, but rather to build on the work of
others. Many past publications are used in this study to validate deformation levels, material
parameters, and response quantities. Agreements between our results and those of other in-
vestigators further establish the dependability of the proposed procedure, while discrepancies
in response quantities have been the motivation for the additional studies presented herein.
This paper essentially builds on and extends the work of Wilson and Tan [6], Werner [1],
Goel and Chopra [2], McCallen and Romstad [7] and Makris et al. [8].
The work of S. D. Werner and others on two-span short bridge overcrossings was summa-
rized in a paper in an eort to evaluate Caltrans procedures for seismic response analysis of
free-way overcrossings [1]. That study underlined the signicance of soilstructure interaction
and oered selected recommended values on some modal parameters. Emphasis was given
to the transverse response of the bridge. As in the Wilson and Tan study [9], the Werner
study [1] did not provide any information on the embankment stinesses and damping along
the longitudinal direction. Information on modal response quantities was limited to the rst
transverse mode only. Despite its limitations and occasional sweeping statements, the Werner
study [1] identies several of the challenges associated with this problem. This study revisits
the Werner paper [1] by rening and extending several of the concepts advanced therein.
A comprehensive study that established the validity of the stick model was conducted by
McCallen and Romstad [7] who computed natural frequencies, mode shapes, and response time
histories of the Painter Street Overcrossing using the two aforementioned numerical models.
Both xed and resilient foundation supports were considered and it was found that there is a
one-to-one correspondence between the mode shapes predicted by the stick and the detailed
nite element model. The McCallen and Romstad study indicated the substantial reduction in
Copyright ? 2002 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2002; 31:19671991
HIGHWAY OVERCROSSINGS 1969
the transverse and longitudinal frequencies when realistic soil strains are considered; while in
these two modes of vibration modal damping should be of the order of 2030%. The levels
of modal damping were concluded by McCallen and Romstad [7] after a large number of trial
and error iterations and comparisons of measured and computed responses at several bridge
locations. The detailed nite element study by McCallen and Romstad [7] that accounts for
the resilience and dissipation at the centre bent and end abutment involved the discretization
of embankments and a large volume of the surrounding soil. Eorts to establish the validity
of the stick model in estimating the seismic response of skew bridges have been also reported
by Werner [1].
Established values of kinematic response functions and dynamic stinesses of the approach
embankments and pile foundations are used in this study to synthesize a simple dynamic model
to estimate the dynamic response of freeway overcrossings. The numerical simulations that
employ a simple stick model and a more sophisticated nite element model build on the work
presented by McCallen and Romstad [7]. In contrast with the McCallen and Romstad study,
the methodology presented in this paper adopts the substructure approach, where the kine-
matic response functions and dynamic stinesses are computed separately and subsequently
are incorporated in a simple dynamic model where the mechanical behaviour of each of its
components can be calculated with any desired level of sophistication. The methodology to
compute the kinematic response functions and dynamic stinesses of approach embankments
is presented in the companion paper, whereas the methodology to compute the dynamic sti-
nesses of pile groups has been summarized in the paper by Makris et al. [8].
Lastly, the structural characteristics that we compute for the Painter Street Overcrossing are
compared with those reported by Goel and Chopra [2], who employed an equilibrium based
approach to back-gure abutment stinesses at dierent levels of shaking. The follow-up work
of Goel [9] is also used to compare the estimated values of modal periods and damping ratios
of the entire bridgefoundation system.
SOILSTRUCTURE INTERACTION
Figure 1(a) shows the schematic of a typical two-span overcrossing with its approach em-
bankments. During ground shaking the dynamic response of the deck is aected: (a) from
the dynamic response of the embankments that support the end abutment and (b) from the
dynamic response of the pile foundations at the centre bent. The need to account for this
interaction motivated nite element studies that involved the discretization of the bridge super-
structure and a large volume of the embankments and supporting soil ([10; 7] among others).
Although such studies elucidated the signicance of deckabutmentembankment interaction,
they do not provide direct information on the distinct mechanical characteristics of approach
embankments and pile foundations and their inuence on the dynamic response of the bridge
structure. This might be a possible reason for the lack of practical procedures to account
for soilstructure interaction when computing the seismic response of freeway overcrossings.
This paper, together with the companion paper [11], concentrates on addressing the issue of
the importance of soilstructure interaction on the seismic response of freeway overcrossings
and proposes a practical methodology to include its eect in association with simple bridge
models.
(a) Vla Real System Vra

la ra
Vf Ura
U la
f
Uf
(b) Kinematic-Seismic Response
[ K la ] [ K ra ]
[ Kf ]
(c) Dynamic Stiffnesses of Pile Group and Embankments

z
[ K la ] [ K ra ]
U la U ra
z
la ra
>0 Vla [ Kf ] V ra
>0 y Uf
f
x Vf
(d) Idealization of Soil-Bridge Interaction
Pile Foundation
x
Protective
U la Systems U ra
y
Protective
Approach
Systems Approach
Embankment
and Pile Embankment
Foundation at and Pile
Abutment Pile Foundation Foundation at
Abutment
(e) Plan View of Idealized Model
Figure 1. General procedure for seismic soilfoundation superstructure interaction.
Owing to its computational eciency the substructure method is a popular approach to

address the soilfoundationsuperstructure problem [12]. Assuming linear soilfoundation
superstructure response, the analysis of the system can be performed in three consecutive
steps as shown in Figure 1. First, nd the motion at the end abutments and pile cap of
the centre bent in the absence of the bridge superstructure (the so-called foundation input
motion), which includes translational as well as rotational components; second, determine the
dynamic stinesses (frequency dependent springs and dashpots) associated with longitudinal,
transverse, vertical, rocking and cross-horizontal-rocking oscillations of the embankments and
pile groups; third, compute the seismic response of the superstructure (deck and abutments)
supported on springs and dashpots and subjected to the foundation input motion.
An earlier attempt to investigate the eect of soilpilestructure interaction was presented
by Makris et al. [8]. In that study the Painter Street Overcrossing was idealized with a
plane six-degree-of-freedom lumped-parameter model, whereas the inuence of the approach
embankments was neglected. The limitations of the plane model restricted the analysis of the
response only along the transverse direction. Despite its limitations, that study indicates some
of the shortcomings that may result by neglecting the resilience of the pile foundations at
the centre bent and outlines a simple integrated procedure that one can follow in order to
compute the stinesses and damping of pile foundations. The same procedure was adopted and
extended in this study which examines the two-dimensional coupled longitudinal and transverse
response of highway overcrossings. Figure 1 (bottom) shows the elevation and plan views of
the model adopted in this study and the springs and dashpots that are proposed to approximate
the interaction of the bridge superstructure with its foundation and the surrounding soil.
Dynamic stinesses of approach embankments
The dynamic stinesses of approach embankments have been examined in the companion
paper and for most practical purposes they can be replaced with frequency independent springs
and dashpots.
Dynamic stinesses of pile groups
The dynamic stinesses of a pile group, in any vibration mode, can be computed using the
dynamic stinesses of a single pile in conjunction with the concept of superposition criterion,
originally developed for static loads by Poulos [13], and later justied for dynamic loads
by Kaynia and Kausel [14], Sanchez-Salinero [15] and Roesset [16]. It can be used with
condence at least for groups with less than 50 piles. Dynamic interaction factors for various
modes of loading are available in the form of non-dimensional graphs [17] and in some cases,
closed form expressions derived from a beam on winkler foundation model in conjunction
with simplied wave-propagation theory [18; 19].
Although the dynamic stinesses of pile groups are in general frequency-dependent quan-
tities, their low frequency values do not uctuate appreciably with frequency and one can
replace them with frequency independent springs and dashpots [11]. The stiness matrix of
a pile foundation is symmetric so that Kx =Kx =moment at the pile cap for a positive unit
displacement along the x direction. With the system of axis shown in Figure 1, a positive
unit displacement along the x direction generates a negative moment. In contrast, a positive
unit displacement in the y direction generates a positive moment so that Ky =Ky 0. K is
the moment along the direction due to a unit rotation in this direction.
RESPONSE OF THE MELOLAND ROAD OVERCROSSING
Figure 2 shows the stick model (left) and the three-dimensional nite element model (right)
of the Meloland Road Overcrossing in its deformed conguration. The stick model is a col-
lection of beam elements with cross-section properties adjusted from geometric data without
f1 = 2.01 Hz
1 = 18.7% f1 = 1.96 Hz
f2 = 2.85 Hz
2 = 56.8% f2 = 2.55 Hz
f3 = 3.58 Hz
3 = 8.3% f3 = 3.41 Hz
f4 = 4.00 Hz f4 = 3.90 Hz
f5 = 4.03 Hz
5 = 28.2% f5 = 4.30 Hz
3 f6 = 4.22 Hz 3
2
1 6 = 10.2% 2
1
f6 = 4.89 Hz
Figure 2. First six natural frequencies and modes computed by stick model (left) and 3D FEM model
(right) of Meloland Road Overcrossing.
considering any cracked section reduction. The three-dimensional nite element model uses
eight-node solid elements for the bridge superstructure. The bridge superstructure is sup-
ported at its centre bent and at each end by the springs and dashpots schematically shown in
Figure 1. The spring and dashpot values of the approach embankments have been estimated
with the methodology presented in the companion paper, whereas the spring and dashpot val-
ues of the pile foundations have been estimated with the methodology presented in the paper
Table I. Spring and dashpot values that approximate the presence of the approach embankments and
pile foundation of the Meloland Road Overcrossing. Values from this study are associated with the
intensity of the 1979 Imperial Valley earthquake.
Parameters 1 2 3 4 5 6
Embankment Kx (MN=m) 21+56 160 91 (365) 107 607+49
+pile foundations (51+49)
Ky (MN=m) 21+56 91 (365) 596+49
Kz (MN=m) 78+356 418 263 (1051)
Cx (MN s=m) 1.5+4.5
Cy (MN s=m) 1.5+4.5
Cz (MN s=m) 3+28
Pile foundation Kx ; Ky (MN=m) 260 254 (876) 1007 175

of centre bent K (MN m=rad) 7611 366 1888 (6509) 5696
Kx (MN=rad) 409
Ky (MN=rad) 409
Kz (MN=m) 887 550 (1898) 1460
Cx ; Cy (MN s=m) 6
Cz (MN s=m) 25
Note: 1. This study (G = 2:0 MPa and = 0:52 for embankment soil).
2. Wilson and Tan [20] ( embankment only, no piles and G = 7:2 MPa is used for embankment soil).
3. Douglas et al. [21] (values in parenthesis are the optimal values identied from dynamic tests).
4. Werner [1] (xed boundary condition at the base of pier).
5. Maragakis et al. [22] (values are from dynamic tests).
6. Caltrans: Method A (Method B) [23].
by Makris et al. [8] and is also included in the report by Zhang and Makris [11]. All values
of interest are summarized in Table I. The loops shown at the far ends of the bridge models
in Figures 2 and 8 represent the deformed embankment springs as being plotted by ABAQUS.
During the numerical simulation, the Youngs modulus of the beam elements on top of
the column was articially increased by three orders of magnitude to form a rigid link in
order to prevent excessive deections at the connection point between pier and deck in the
stick model [7]. Vertical excitations are not considered. In both models, the damping of the
bridge deck and centre bent is approximated with the Rayleigh damping approximation, where
the parameters and are computed by assuming a 5% modal damping ratio in the rst
and the second modes. The Youngs modulus of the concrete is assumed to be Ec =22 GPa.
This value is approximately 80% of the value obtained from empirical expressions to account
for the cracking that occurred during the earthquake. Similar cracked values for the Youngs
modulus of concrete in seismic response analysis of bridges have been reported by Douglas
and Reid ([24, Ec =2025 GPa]) and Dendrou et al. ([25, Ec =20 GPa]). The density of
concrete is assumed 2400 kg=m3 .
Eigenvalue analysis
Eigensolutions were performed for the stick model and three-dimensional nite element model
using the commercially available software ABAQUS. Figure 2 compares the rst six modes
and modal frequencies of the stick model and three-dimensional nite element model, and
indicates a one-to-one correspondence between the two models. The natural frequencies of
the stick model are also in good agreement with that of the three-dimensional nite element
model.
While modal frequencies are directly estimated by solving the real eigenvalue problem
of some structural idealization of the bridgefoundation system having mass matrix, [M ],
and stiness matrix, [K], the estimation of the modal damping ratios appears to be a less
straightforward procedure. The majority of modal damping ratios of bridges published in the
literature have been back-gured by processing recorded data with system identication algo-
rithms (Wilson [26], Werner et al. [4]; Wilson and Tan [6]; Werner [1]; and Goel [9] among
others). Although more recent multi-inputmulti-output system identication algorithms ap-
pear to be more ecient than older single-inputsingle-output algorithms, the relevance of the
reported values is strongly associated with the sophistication of the adopted structural model
and the quality of the recorded data. In some occasions modal damping values appreciably
larger than those identied were recommended [1].
In an eort to calibrate nite element results, McCallen and Romstad [7] initially assigned a
uniform modal damping ratio, j =5%, to the rst six modes of the Painter Street Overcrossing.
Subsequently, after conducting a large number of trial and error iterations and comparisons of
recorded and computed responses of the bridge, they reassigned much larger values of modal
damping to selected modes in order to approximate satisfactorily the recorded responses. In
this study the modal damping ratios of the bridge foundation system are computed by solving
for the complex eigenvalues of the homogeneous equation
[M ]{u } + [C]{u} + [K]{u} =0 (1)
where [M ]; [C], and [K] are the mass, damping, and stiness matrices of the bridgefoundation
idealization shown in Figure 1 and {u} is the free vibration response vector
{u} = {}eit (2)
In Equation (2), is the complex characteristic value and {} is the associated characteristic
vector. This complex eigenvalue approach is well known in the literature and was given a
critical review by Veletsos and Ventura [27]. The damping matrix, [C], is constructed by
adopting the concept of Rayleigh damping for the bridge superstructure and appending the
pre-identied lumped dashpots at the locations where the superstructure interacts with its
foundation. Following this approach we assign a 5% modal damping ratio at the rst and
second modes of the undamped idealized model (bridge deck with springs) and add the
damping constants, cij , identied in the companion paper ([11], for the embankments) and
in the paper by Makris et al. ([8], for the pile groups), that represent the presence of the
embankments and pile foundations. With this superposition the non-classical damping matrix,
[C], of the bridgefoundation system assumes the form
[C]=[M ] + [K] + [cij ] (3)
The matrix [cij ] is assembled in the same way as the stiness matrix [K] of the undamped
superstructure so that the lumped dashpots cij are assigned to the correct degrees of freedom.
Substitution of Equation (2) in Equation (1) yields
(2 [M ] + i[C] + [K]){} =0 (4)
which is a standard polynomial eigenvalue problem. The roots of this polynomial are the
complex characteristic values j that were evaluated with MATLAB [28]. The relation be-
tween j with the modal frequencies, !j , and modal damping ratios, j (j =1; 2; : : : ; N ) is
determined from the associated equation of a single-degree-of-freedom oscillator. Interpreting
j as the frequency domain parameter one may determine

j = !j2 2j !j2 + ij !j (5)
After solving for !j and j ,

!j = 2jR + 2jl (6)
jl
j = (7)
!j
in which jR and jl are the real and imaginary parts of the characteristic value j , respec-
tively.
The stick model used in the real eigenvalue analysis consists of 354 degrees-of-freedom for
Meloland Road Overcrossing. In order to bypass the problem of computing and interpreting the
complex eigenvalues of such large matrices, a reduced-order stick model was developed with
fewer degrees of freedom. Sensitivity studies indicated that the modal characteristics of the
rst ten modes are virtually insensitive to the exact number of elements of the reduced-order
stick model. Two reduced-order stick models that consist of 228 and 300 degrees-of-freedom
(d.o.f), respectively yield nearly identical results as indicated in Table II.
Table II presents the rst six real eigenvalues of the Meloland Road Overcrossing that
have been computed with the 3D nite element model of the undamped bridge (column A),
the original stick model of the undamped bridge (column B, d:o:f =354), the reduced-order
stick model of the undamped bridge (column C, d:o:f =300 and 228, respectively), and the
rst six complex eigenvalues of the reduced-order stick model of the damped bridge (column
D, d:o:f =300 and 228, respectively). The corresponding damping ratios computed with this
study are shown in column 1, next to other values reported in the literature.
The values shown in Table II indicate that the reduced-order stick model in association with
the complex eigenvalue analysis yield valuable information on the free vibration characteristics
of the bridge.
The computed modal damping ratios, j , are much larger than the 5% modal damping
assumed by Caltrans.
The computed rst modal damping ratio, 1 =18:7%, which in this case corresponds to
the rst transverse mode with the deck bending away from the undeformed conguration,
is in agreement with the low end of damping values along the transverse mode that are
reported by Werner [1]. The values of rst modal damping reported by Werner [1] are
appreciably larger than the value by the same and other investigators during earlier studies
on the same bridge [4].
The computed second modal damping 2 =56:8% is unusually high. However, our con-
dence in this value originates from the straight conguration of the deck and its integral
1976
Table II. Modal frequencies, !j (rad=s), and damping ratios, j (%), of the Meloland Road Overcrossing.
Modes Eigenvalues (rad= s) 1 2 3 4 5

Model A Model B Model C Model D !j j !j j !j j !j !j j
First transverse 12.320 12.647 12.440 13:519 + 2:5771i 13.8 18.7 15.5 7.2 14.315.7 6.612.7 15.6 16.3 1926
(12.441) (13:702 + 2:7333i) (13.8) (18.7)
Longitudinal 16.044 17.913 17.882 16:001 + 11:047i 19.4 56.8 16.7
Copyright ? 2002 John Wiley & Sons, Ltd.

(17.884) (16:221 + 11:104i) (19.4) (56.8)
First vertical 21.405 22.460 21.911 21:073 + 1:746i 21.1 8.3 17.5
(antisymmetric) (21.921) (21:252 + 1:888i) (21.2) (8.3)
Torsion about 24.521 25.105 23.823 0:0 + 17:942i 17.9 100 Critically damped mode
vertical axis (23.824) (0:0 + 17:962i) (18.0) (100)
Second transverse= 26.966 25.340 25.096 25:778 + 7:570i 26.9 28.2

torsion
J. ZHANG AND N. MAKRIS
about longitudinal (25.087) (26:830 + 8:010i) (26.9) (28.2)

axis
Second vertical 30.724 26.515 26.496 28:153 + 2:895i 28.3 10.2 28.7 5.8 27.429.9 3.17.4 27.6
(symmetric) (26.526) (28:187 + 2:900i) (28.3) (10.2)
Note: A: Undamped original 3D FEM model.

B: Undamped original stick model.
C: Undamped reduced stick model with 300 d.o.f and (228 d.o.f), respectively.
D: Damped reduced stick model with 300 d.o.f and (228 d.o.f), respectively.
1: This study.
2: Werner et al. [4].
3: Wilson and Tan [6].
4: Gates [29].
5: Werner [1].
Earthquake Engng Struct. Dyn. 2002; 31:19671991

abutments which are mobilizing a large volume of soil with high damping. Indirect
evidence that this high damping value might be realistic is provided by the inability of
system identication studies to detect modal characteristics along the longitudinal mode
of vibration.
The modal damping associated with the third mode (1st vertical, antisymmetric, 3 =8:3%)
is lower than the modal damping associated with the transverse and the longitudinal modes
that involve more soil participation.
The torsional mode about the vertical axis is critically damped. Since the modes are
decoupled the modal damping of the torsional mode can be estimated by assuming that
the deck is a single-degree-of-freedom structure with length, L, and linear mass density,
m, that rotates about the centre bent and its motion is resisted at the two ends by the
transverse springs and dashpots of the embankments indicated in Table I. With this
idealization, the equation of motion for torsion about the vertical axis is
I0 + 12 Cx L2 + 12 Kx L2 =0 (8)
where I0 = 121 mL3 is the moment of inertia of the deck about the vertical axis. Using the
standard real valued procedure [30]
or the complex formulation given by Equations (6)
and (7) one concludes that !5 = 6Kx =mL=26:2 rad=s and 5 =(3Cx )=( 6Kx mL)1:0.
The computed fth modal damping 5 =28:6% also assumes a high value, probably due
to the explanation oered above, since this mode involves the transverse motion of the
deck as well as torsional motion of deck about the longitudinal axis.
The good agreement between the values reported for the rst and sixth natural frequencies
of the MRO by various investigators is worth mentioning.
Time history analysis
Figures 35 plot total acceleration, relative velocity, and displacement time histories of the
bridge response at selected locations. The analysis shown in these gures investigates the
sensitivity of the bridge response to the support motion. The rst column shows the recorded
motions. The second column (case 1) shows computed response quantities by using as a
support motion at the end abutments the crest motions computed using Equation (14) in
the companion paper. The third column (case 2) shows the computed response quantities by
using as a support motion at the centre bent and the end abutments the free-eld motion.
The last column (case 3) shows computed response quantities by using as a support motion
at the end abutments the recorded crest motions. All simulated responses are obtained by
implementing the spring and dashpot values that have been evaluated with the approximate
methods advanced in the companion paper.
Figure 3 compares the computed response with the records of channel 7, which is located
at mid-span of the bridge deck (see Figure 6 of the companion paper). The case where the
recorded crest motion are used as support motions yields the best overall predictions. The
acceleration and the deck drift are predicted with errors less than 8%, while the relative
velocity is underestimated by 26%. When the computed crest motions are used as support
motions, the peak acceleration is overestimated by 27% while the deck drift and the relative
velocity is predicted with marginal discrepancies. Figure 4 compares the computed response
with the records of channel 8. When the free-eld motions are used as support motions the
Recorded - Ch. 7 Computed (case 1) Computed (case 2) Computed (case 3)

0.5 0.5 0.5 0.5
A (g)
0 0 0 0
_0.5 _0.5 =27.2% _0.5 =17.2% _0.5 =8.1%
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
=16.4% =61.1% =26.3%

_0.5 _0.5 _0.5 _0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
_0.05 _0.05 _0.05 _0.05

=5.3% =69% =5.6%
0 5 10 0 5 10 0 5 10 0 5 10
Time (s) Time (s) Time (s) Time (s)
Figure 3. Records of channel 7 and predictions of Meloland Road Overcrossing response

considering dierent support motions.
discrepancies between records and predictions are of the order of 70% or more for relative
velocities and relative displacements. When the computed crest motions are used as support
motions the peak acceleration is overestimated by 50% while the deck drift history contains
higher frequencies than the recorded. Similar trends can be observed in Figure 5, which
compare the computed responses with the records of channel 13. The results shown in Figures
35 indicate the signicance of considering the amplied support motions at the crest of the
embankments. More comparisons between recorded and computed time histories are oered
in the report by Zhang and Makris [11].
Our parametric analysis proceeds by analysing the bridge response when dierent support
idealizations are considered. Figure 6 shows three idealizations of interest: (a) monolithic
abutments and viscoelastic foundation at the centre bent, (b) viscoelastic embankments and
monolithic support at the centre bent, and (c) viscoelastic embankments and elastic support at
the centre bent. The sensitivity of the bridge responses to the resilience and dissipation of the
bridge supports is investigated in Figure 7, which compares the response quantities against
the records of channel 7. The rst column in Figure 7 shows the recorded motions. The
second column shows computed response quantities by assuming monolithic abutments and
viscoelastic foundation at the centre bent. It examines the eect of neglecting the resilience and
dissipation at the abutments. It is shown that in this case the response is more high frequency
with overestimated peak acceleration. The third column examines the eect of neglecting the
resilience and dissipation of the pile foundation at the centre bent. The last column examines

0.5 0.5 0.5 0.5
A (g)
0 0 0 0
=50.5% =4.7% =11.6%

0.5 0.5 0.5 0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
=56.8% =71.1% =26.7%

0.5 0.5 0.5 0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.1 0.1 0.1 0.1

Rel. U (m)
0.05 0.05 0.05 0.05
0 0 0 0
0.05 0.05 0.05 0.05

=23% =89.4% =64.1%
0.1 0.1 0.1 0.1
0 5 10 0 5 10 0 5 10 0 5 10

the eect of neglecting the dissipation of the pile foundation at the centre bent. All simulated
response are subjected to the recorded motions at the crest of the embankment and the free-
eld and should be compared with the last column of Figure 3. The case of a rigid support
at the centre bent (third column) results in the smaller drifts, while the case of neglecting the
dissipation of the pile foundations [2] yields comparable results when it is included.
RESPONSE OF THE PAINTER STREET OVERCROSSING
Figure 8 shows the stick model (left) and three-dimensional nite element model (right)
of the Painter Street Overcrossing in its deformed congurations. Again, the stick model
is a collection of beam elements with cross-section properties adjusted from geometric data
without considering any cracked section reduction. While it is not shown in Figure 8 the beam
elements are joined following the skew conguration of the bridge that results in coupling
of the vibration modes. The three-dimensional nite element model uses eight-node solid
elements for the bridge structure. The bridge superstructure is supported at its centre bent and
both ends by the springs and dashpots schematically shown in Figure 1their values being
estimated separately in the companion paper and summarized in Table III (column 1), along
with selected values reported in literature (columns 25).

0.5 0.5 0.5 0.5
A (g)
0 0 0 0
=10.2% =30.1% =18.5%

0.5 0.5 0.5 0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
=25.4% =75.9% =26.8%

0.5 0.5 0.5 0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
0.05 0.05 =15.1%

0.05 =78.9%
0.05 =9.5%
0 5 10 0 5 10 0 5 10 0 5 10

Vertical excitations are not included during the analysis. In both models, the damping of
the bridge superstructure is approximated with the Rayleigh damping approximation, where
the parameters and are computed by assuming a 5% modal damping ratio in the rst and the
second modes. The Youngs modulus of the concrete is assumed to be 22 GPa. This value
is approximately 80% of the value obtained from empirical expressions to account for the
cracking that occurred during the earthquake. The density of concrete is 2400 kg=m3 .
Eigenvalue analysis
Eigensolutions were performed for the stick model and three-dimensional nite element model
using the commercially available software ABAQUS. Figure 8 compares the rst six natural
modes and frequencies of the stick model and the three-dimensional nite element model.
There is a one-to-one correspondence between the mode shapes predicted by the stick model
and three-dimensional nite element model. The natural frequencies of the stick model are
also in good agreement with that of the three-dimensional nite element model. The rst
values shown are those computed by adopting the converged soil properties during the strong
1992 Petrolia earthquake. The values shown in the parenthesis are those reported by McCallen
and Romstad [7]. They are 50% higher than the values computed in this study. Part of the
reason for these discrepancies is the six times larger embankment stinesses they used in their
study.
recorded
recorded crest motions
crest motions recorded
crest motions
free-field motions
free-field motions
Monolithic embankments and viscoelastic Viscoelastic embankments and monolithic

(a) foundation at the center bent (b) support at the center bent
recorded crest motions recorded crest motions
free-field motions
(c) Viscoelastic embankments and elastic support at the center bent
Figure 6. Bridge models with dierent support idealizations.
Modal damping ratios are estimated with the complex eigenvalue procedure presented in
the section associated with the analysis of the Meloland Road Overcrossing. Similarly, a
reduced-order stick model was developed with fewer degrees-of freedom in order to bypass
the problem of computing and interpreting the large number of complex eigenvalues resulting
from the original stick model which consists of 618 degrees of freedom. Table IV presents
the rst six eigenvalues of the Painter Street Overcrossing that have been computed with the
three-dimensional nite element model of the undamped bridge (column A), the original stick
model of the undamped bridge (column B), the reduced-order stick model of the undamped
bridge (column C); and the rst six complex eigenvalues of the reduced-order stick model of
the damped bridge. Damping ratios computed with this study are shown in column 1 next to
other values reported in the literature. Selected observations from the modal values indicated
in Table IV are:
Due to the skew conguration, there is strong coupling of modes involved in the eigen-
value analysis.
The computed modal damping ratio, j , are larger than the 5% modal damping assumed
by Caltrans, however, the rst three modal damping ratios are not as high as the modal
damping ratios computed for the straight (unskewed) Meloland Road Overcrossing. This
might be partly due to the strong participation of the vertical mode in the Painter Street
Bridge that is associated with less damping.
The computed rst modal damping, l =9% is approximately half the value that McCallen
and Romstad [7] needed to match the recorded data. However, it is in agreement with
the high-end of the modal damping range identied by Goel [9].

0.5 0.5 0.5 0.5
A (g)
0 0 0 0
0.5 0.5 =16.8% 0.5 =1.29% 0.5 =9.92%
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
=5.64% =39.1% =25.7%

0.5 0.5 0.5 0.5
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
0.05 0.05 =4.63%

0.05 =27.4%
0.05 =5.07%
0 5 10 0 5 10 0 5 10 0 5 10

considering dierent support idealizations.
The longitudinal mode (that was the second mode for the Meloland Road Overcrossing)
has been pushed down to the fth mode. Interestingly, the complex eigenvalue analysis
advanced in this study is able to capture the high damping, 5 =46%, associated with
this mode. As was indicated during the analysis of the modal properties of the Meloland
Road Overcrossing, in which 2 =56:5%, the longitudinal mode mobilize a large volume
of soil with high damping.
The procedure advanced in this study, where the appropriate values of G and of the soil
embankments are established with the kinematic response analysis, yields a rst natural
frequency that is in very good agreement with the values reported by Goel [9] and Price
and Eberhard [31].
Time history analysis
Figures 9 and 10 plot total acceleration, relative velocity, and displacement time histories of
the bridge response at selected locations. The analysis shown in these gures investigates the
sensitivity of the bridge response to the foundation input motion. The rst column shows the
recorded motions. The second column (case 1) shows computed response quantities by using
as a support motion at the end abutments the crest motions computed using Equation (14)
of the companion paper. The third column (case 2) shows the computed response quantities
by using as a support motion at the centre bent and end abutments the free-eld motion. The
f1 = 1.78 Hz ( 2.69 Hz)

1 = 9.0% f1 = 2.31 Hz ( 3.06 Hz)
f2 = 2.29 Hz ( 3.30 Hz)

2 = 6.7% f2 = 2.80 Hz ( 4.29 Hz )
f3 = 2.60 Hz ( 4.00 Hz) f3 = 2.93 Hz ( 4.45 Hz )

3 = 5.8%

, 5 Hz ( 4.72 Hz)
f4 = 32
4 = 8.3% f4 = 3.75 Hz ( 5.03 Hz)
f5 = 3.29 Hz ( 5.24 Hz)

5 = 45.8% f5 = 4.24 Hz ( 5.28 Hz)
f6 = 3.43 Hz ( 6.61 Hz)

3
6 = 17.0% 2 f6 = 5.13 Hz ( 5.53 Hz)
1
Figure 8. First six natural frequencies and modes computed by stick model (left) and
3D FEM model (right) of the Painter Street Overcrossing. Values in parentheses are
those reported by McCallen and Romstad [7].
last column (case 3) shows computed response quantities by using as a support motion at the
end abutments the recorded crest motions.
Figure 9 compares the computed responses with the records of channel 4. The case where
the recorded crest motions are used as support motions (last column) yields invariably the
most favorable prediction. When the free-eld motions are used as support motions, the bridge
Table III. Spring and dashpot values that approximate the presence of the embankments and pile foun-
dations of the Painter Street Overcrossing. Values from this study are associated with the intensity of
the 1992 Petrolia earthquake.
Parameters 1 2 3 4 5
Embankment Kx (MN=m) 137+180 201 851+105 117 438+ 810+105
+pile foundations (68+105)
Ky (MN=m) 137+180 815+105 1461458+ 876+105
Kz (MN=m) 582+773 564
Cx (MN s=m) 9+9
Cy (MN s=m) 9+9
Cz (MN s=m) 17+56
Pile foundation Kx ; Ky (MN=m) 321 140 140

of centre bent K (MN m=rad) 5254
Kx (MN=rad) 354
Ky (MN=rad) 354
Kz (MN=m) 982
Cx ; Cy (MN s=m) 5
Cz (MN s=m) 20
Note: 1. This study (G = 8:0 MPa and = 0:50 for embankment soil).
2. Wilson and Tan [20] ( embankment only, no piles).
3. McCallen and Romstad [7].
4. Goel and Chopra [2] (+ embankment and piles).
5. Caltrans Method A (Method B) [23].
response shown along the third column is substantially underestimated, since it has not expe-
rienced the amplication that the embankments induce at the two ends. When the computed
crest motions are used as support motions the peak accelerations are predicted with marginal
discrepancies; however, deck drifts are underestimated by 22% with a time history that ex-
hibits higher frequencies than the recorded. Figure 10 which compares the computed response
with the records of channel 7, indicates similar trends. When the free-eld motions are used
as support motions the discrepancies between records and predictions are of the order of 40%
or more, for relative velocities and relative displacements. When the computed crest motions
are used as input, the computed responses are as good as when the recorded crest motions
are used. Additional comparisons between recorded and computed time histories are oered
in the report by Zhang and Makris [32].
The sensitivity of the bridge response to the resilience and dissipation of the bridge supports
is investigated in Figures 11 and 12, which plot total acceleration, relative velocity, and
displacement time histories. Again the rst column shows the recorded motion for convenience.
The second column plots the computed response quantities by assuming that the abutments
are monolithic supports and soilstructure interaction happens only at the foundation of the
centre bent (case (a) of Figure 6). The third column plots the computed response quantities
by considering soilstructure interaction at the end abutments and by assuming a monolithic
support at the centre bent (case (b) of Figure 6). The fourth column plots the computed
response quantities by neglecting the dissipation of the pile foundation at the centre bent. In
all cases the free-eld motions were induced at the foundation of the centre bent and the
recorded crest motions were induced at the end abutments.
Table IV. Modal frequencies, !j (rad=s), and damping ratios, j (%), of the Painter Street Overcrossing.
Modes Eigenvalues (rad=s) 1 2 3 4

Model A Model B Model C Model D !j j !j j !j j !j j
First transverse= 14.514 11.162 11.364 11:490 + 1:040i 11.5 9.0 20.7 20 11.017.9 5.68.5 10.3 16.6
antisymmetric vertical (11.587) (11:730 + 1:116i) (11.8) (9.5)
Copyright ? 2002 John Wiley & Sons, Ltd.

Antisymmetric vertical= 17.593 14.409 14.578 14:683 + 0:984i 14.7 6.7 16.9 3
torsion about vertical axis (14.751) (14:863 + 1:006i) (14.9) (6.8)
Torsion about vertical 18.410 16.366 16.365 16:527 + 0:962i 16.6 5.8 25.1 3
axis= symmetric vertical (16.366) (16:524 + 0:972i) (16.6) (5.9)
Symmetric vertical= 23.562 20.691 20.808 21:075 + 1:761i 21.1 8.3 32.9 5
longitudinal (20.994) (21:370 + 1:809i) (21.4) (8.4)
Longitudinal 26.641 21.545 20.938 20:176 + 10:383i 22.7 45.8 29.6 30

(21.265) (20:394 + 10:395i) (22.9) (45.4)
HIGHWAY OVERCROSSINGS
Second transverse=torsion 32.233 31.156 22.052 23:754 + 4:096i 24.1 17.0 41.5 5
about longitudinal axis (22.532) (24:236 + 4:302i) (24.6) (17.5)
Note: A: Undamped original 3D FEM model.
B: Undamped original stick model.
C: Undamped reduced stick model with 174 d.o.f and (138 d.o.f), respectively.
D: Damped reduced stick model with 174 d.o.f and (138 d.o.f), respectively.
1: This study.
2: McCallen and Romstad [7].
3: Goel [9].
4: Price and Eberhard [31].
Earthquake Engng Struct. Dyn. 2002; 31:19671991

1985

1 1 1 1
A (g) 0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 0.5 0.5
1 1 =14.4% 1 =57.3% 1 =3.9%
0 5 10 0 5 10 0 5 10 0 5 10
Rel. V (m/s
0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 =15.5%

0.5 =59.5%
0.5 =12.6%
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
0.05 0.05 =21.5%

0.05 =62.4%
0.05 =4.6%
0 5 10 0 5 10 0 5 10 0 5 10
Figure 9. Records of channel 4 and predictions of Painter Street Overcrossing response

Figure 11 compares the computed responses with the records of channel 4. It indicates
that the exibility of the pile foundation of the centre bent is appreciably aecting the re-
sponse, whereas the associated damping has less important eects. Neglecting the exibility
and damping of the end abutments results to a more high-frequency acceleration response.
Figure 12, which shows computed responses at the mid-span, indicates the similar trends.
More specically, the results obtained by neglecting the damping of the pile foundation at
the centre bent are comparable to the results obtained when damping is included, a result
that conrms the validity of the assumption adopted by Goel and Chopra [2]. These trends
suggest that in this case the damping of the pile foundation at the centre bent has marginal
eects; whereas the exibility of the pile foundations has an appreciable eect and has to be
included.
OUTLINE OF PROPOSED PROCEDURE
The two case studies presented in this paper conrmed the validity of a step-by-step procedure
to estimate the seismic response of freeway overcrossings. The study shows that the stick
model used by Caltrans when enhanced with realistic springs and dashpots at its supports can
yield dependable estimates of the seismic response of freeway overcrossings.

1 1 1 1
A (g) 0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 0.5 0.5

=9.4% =2.3% =8.1%
1 1 1 1
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
0.5 0.5 =5%

0.5 =43.8%
0.5 =4.9%
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
=15.7% =61% =18%

0.05 0.05 0.05 0.05
0 5 10 0 5 10 0 5 10 0 5 10
Step 1: Compute the kinematic response function of the approach embankments as outlined
in the companion paper, after establishing the converged values of the equivalent linear soil
parameters, G and .
Step 2: Compute the embankment crest response by amplifying the free-eld motion with
the kinematic response functions obtained in step 1.
Step 3: Compute the frequency-independent spring and dashpot values of the approach
embankment using the values of G and established in step 1, as shown in the companion
paper.
Step 4: Compute the frequency-independent spring and dashpot values of the pile groups
at the abutments and the centre bent as outlined in the paper by Makris et al. [8].
Step 5: Construct a stick model of the bridge enhanced with the transverse and longitudinal
spring and dashpot values computed in steps 3 and 4.
Step 6: Compute the two-dimensional dynamic response of the model constructed in
step 5 (see Figure 1) subjected to the free-eld motions at the centre bent and the crest
motions at the abutment ends computed in step 2.
CONCLUSIONS
The free-vibration and earthquake responses of two instrumented bridges are examined in this
paper with a reduced-order stick model and a more detailed three-dimensional nite element

1 1 1 1
A (g) 0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 0.5 0.5
1 1 =32% 1 =17.2% 1 =8%
0 5 10 0 5 10 0 5 10 0 5 10
Rel. V (m/s
0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 0.5 0.5

=17.4% =26% =20.9%
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
0.05 0.05 =4%

0.05 =24.6%
0.05 =11.7%
0 5 10 0 5 10 0 5 10 0 5 10
model. Either model in this study was enhanced with the springs and dashpots established in
the companion paper to account for the presence of the approach embankments and pile foun-
dations. Our analysis revealed distinguishable trends that lead to the following conclusions:
The reduced-order stick model yields comparable modal parameters and seismic response
characteristics to the more detailed three-dimensional nite element model. It is capable to
capture the dynamic characteristics of a skewed overcrossing that exhibits strong coupling
of its vibration modes.
The modal damping ratios, j , of either the straight Meloland Road Overcrossing and the
skewed Painter Street Overcrossing are much larger than the 5% modal damping ratios
assumed by Caltrans.
The rst mode of the straight Meloland Road Overcrossing is the transverse mode;
whereas, for the skewed Painter Street Overcrossing, it is the coupled transverse=rst
antisymmetric vertical mode.
The rst modal damping ratio of the Meloland Road Overcrossing is of the order of
20%; whereas, for the Painter Street Overcrossing, it is of the order of 10%. The smaller
amount of the rst modal damping in the skewed Painter Street Overcrossing is because
the transverse mode is coupled with the rst antisymmetric vertical mode that is lightly
damped.

1 1 1 1
A (g) 0.5 0.5 0.5 0.5
0 0 0 0
0.5 0.5 0.5 0.5

=17.4% =30.6% =8.9%
1 1 1 1
0 5 10 0 5 10 0 5 10 0 5 10
0.5 0.5 0.5 0.5

Rel. V (m/s
0 0 0 0
0.5 0.5 =4.2%

0.5 =82.4%
0.5 =17.3%
0 5 10 0 5 10 0 5 10 0 5 10
0.05 0.05 0.05 0.05

Rel. U (m)
0 0 0 0
=14.9% =84.4% =9.7%

0.05 0.05 0.05 0.05
0 5 10 0 5 10 0 5 10 0 5 10
The longitudinal mode emerges as the second mode for the straight Meloland Road
Overcrossing; whereas, in the skewed Painter Street Overcrossing, it is pushed down to
the fth place. In both cases the modal damping ratio along the longitudinal direction is
of the order of 50%.
Vertical vibration modes exhibit smaller damping ratios (810%).
The torsional mode of the straight and symmetric overcrossings is highly damped. For
the particular case of the Meloland Road Overcrossing, the torsional mode (4th mode)
was critically damped, 4 100%. In contrast, in the case of the skewed Painter Street
Overcrossing the torsional mode is coupled with the symmetric vertical mode and the
modal damping was as low as 3 6%.
Time history analysis shows that the amplied crest motions of the approach embank-
ments have an appreciable eect on the bridge response and should not be neglected.
When the computed crest motions are used as support motions at the abutments the
prediction of the response is far superior than when one uses the free-eld motions.
Parametric studies that examine the eect of dierent support idealizations indicate that
neglecting the resilience of the foundation of the centre bent yields unrealistic small drifts,
while neglecting the dissipation of the foundation of the centre bent has a marginal eect.
In summary, in view of the strong eect of soilstructure interaction, it is concluded that
the earthquake response of highway overcrossings can be realistically estimated with the
stick model used by Caltrans provided that (a) it is enhanced with the springs and dashpots
established in the companion paper and (b) it is subjected at its end abutments to the amplied
crest motions.
ACKNOWLEDGEMENTS
Partial nancial support for this study was provided by the National Science Foundation under Grant
CMS-9696241. The valuable input and comments of Dr Tim Delis are appreciated.
REFERENCES
1. Werner SD. Study of Caltrans seismic evaluation procedure for short bridges. Proceedings of the 3rd Annual
Seismic Research Workshop. California Department of Transportation, Sacramento, California, 1994.
2. Goel RK, Chopra AK. Evaluation of bridge abutment capacity and stiness during earthquakes. Earthquake
Spectra 1997; 13(1):123.
3. Maragakis EA, Jennings PC. Analytical models for the rigid body motions of skew bridges. Earthquake
Engineering and Structural Dynamics 1987; 15(8):923944.
4. Werner SD, Beck JL, Levine MB. Seismic response evaluation of Meloland road overpass using 1979 Imperial
valley earthquake records. Earthquake Engineering and Structural Dynamics 1987; 15:249274.
5. Crouse CB, Hushmand B, Martin GR. Dynamic soilstructure interaction of single-span bridge. Earthquake
Engineering and Structural Dynamics 1987; 15:711729.
6. Wilson JC, Tan BS. Bridge abutments: accessing their inuence on earthquake response of Meloland Road
Overpass. Journal of Engineering Mechanics 1990; 116(8):18381856.
7. McCallen DB, Romstad KM. Analysis of a skewed short-span, box-girder overpass. Earthquake Spectra 1994;
10(4):729755.
8. Makris N, Badoni D, Delis E, Gazetas G. Prediction of observed bridge response with soilpilestructure
interaction. Journal of Structural Engineering 120(10):29923011.
9. Goel RK. Earthquake behaviour of bridges with integral abutments. Proceedings of the National Seismic
Conference on Bridges and Highways: Progress in Research and Practice. Sacramento, CA, 1997; 149159.
10. Sweet J. A technique for nonlinear soilstructure interaction. Rep. CAI093-100, Sacramento, California, 1993.
11. Zhang J, Makris N. Kinematic response functions and dynamic stinesses of bridge embankments. Earthquake
Engineering and Structural Dynamics (Companion Paper), 2001.
12. Tseng W-S, Penzien J. Soilfoundationstructure interaction. In Chen W-F, Duan L (eds.), Bridge Engineering
Handbook (Chapter 42). CRC Press: Boca Raton, FL, 2000.
13. Poulos HG. Analysis of the settlement of pile groups. Geotechnique 1968; 18(4):449 471.
14. Kaynia AM, Kausel, E. Dynamic stiness and seismic response of sleeved piles. Report No. R80-12, MIT,
Cambridge, MA, 1982.
15. Sanchez-Salinero I. Dynamic stiness of pile groups: approximate solutions. Geotech. Engrg. Report No. GR83-
5, University of Texas, at Austin, TX, 1983.
16. Roesset JM, Angelides D. Dynamic stiness of piles. Numerical Methods in Oshore Piling. Institution of
Civil Engineers: London, 1980; 7582.
17. Gazetas G, Fan K, Kaynia A, Kausel E. Dynamic interaction factors for oating pile group. Journal of
Geotechnical Engineering, ASCE 117(10):15311548.
18. Dobry R, Gazetas G. Simple method for dynamic stiness and damping of oating pile groups. Geotechnique
1988; 38:557574.
19. Makris N, Gazetas G. Dynamic pile-soil-pile interaction Part II: Lateral and seismic response. Earthquake
Engineering and Structural Dynamics 1992; 21(2):145162.
20. Wilson JC, Tan BS. Bridge abutments: formulation of simple model for earthquake response analysis. Journal
of Engineering Mechanics 1990; 116(8):18281837.
21. Douglas BM, Maragakis E, Vrontinos S. Parameter identication studies of the Meloland Road Overcrossing.
Proceedings of the Pacic Conference on Earthquake Engineering, Auckland, New Zealand, vol. 1. 1991;
105116.
22. Maragakis E, Douglas BM, Abdel-Ghaar SM. An equivalent linear nite element approach for the estimation
of pile foundation stinesses. Earthquake Engineering and Structural Dynamics 1994; 23(10):11151124.
23. CALTRANS. Bridge Design Aids 14-1. California Department of Transportation, Sacramento: CA, 1989.
24. Douglas BM, Reid WH. Dynamic tests and system identication of bridge. Journal of the Structural Division,
ASCE 1982; 108(ST10):22952312.
25. Dendrou B, Werner SD, Toridis T. Three-dimensional response of a concrete bridge system to traveling seismic
waves. Computers and Structures 1985; 20:593603.
26. Wilson JC. System identication of the seismic response of a highway bridge. Proceedings of the 8th European
Conference on Earthquake Engineering 1986; 3:1724.
27. Veletsos AS, Ventura CE. Modal analysis of non-classically damped linear systems. Earthquake Engineering
and Structural Dynamics 1986; 14:217243.
28. MATLAB. Mathworks Inc., Natick, MA, 1997.
29. Gates JH. Dynamic eld response studies and earthquake instrumentation of the Meloland Road Overcrossing.
Structural Engineering in Natural Hazards Mitigation: Proceedings of the Papers Presented at the Structures
Congress93, vol. 1. ASCE, New York, NY, 1993; 343348.
30. Chopra AK. Dynamics of Structures: Theory and Applications to Earthquake Engineering. Prentice Hall:
Englewood Clis, NJ, 1995.
31. Price TE, Eberhard MO. Ecient procedure for modeling the transverse seismic response of bridge embankments.
Proceedings of the Sixth U.S. National Conference on Earthquake Engineering, Earthquake Engineering
Research Inst., Oakland, CA, 1998.
32. Zhang J, Makris N. Seismic response analysis of highway overcrossings including soilstructure interaction,
PEER-2001/02, Pacic Earthquake Engineering Research Center, University of California, Berkeley, 2001.

Zhang Et Al-2002-Earthquake Engineering &amp; Structural Dynamics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zhang Et Al-2002-Earthquake Engineering &amp; Structural Dynamics

Uploaded by

Copyright:

Available Formats

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2002; 31:19671991 (DOI: 10.1002/eqe.197)

Seismic response analysis of highway overcrossings including

Jian Zhang and Nicos Makris;

KEY WORDS: dynamic analysis, highway bridges, earthquakes, soilstructure interaction

Contract=grant sponsor: National Science Foundation; contract=grant number: CMS 9696241.

Received 9 May 2001

(a) Vla Real System Vra

(b) Kinematic-Seismic Response

(c) Dynamic Stiffnesses of Pile Group and Embankments

Figure 1. General procedure for seismic soilfoundation superstructure interaction.

Owing to its computational eciency the substructure method is a popular approach to

Dynamic stinesses of approach embankments

Dynamic stinesses of pile groups

RESPONSE OF THE MELOLAND ROAD OVERCROSSING

Pile foundation Kx ; Ky (MN=m) 260 254 (876) 1007 175

After solving for !j and j ,

Modes Eigenvalues (rad= s) 1 2 3 4 5

Longitudinal 16.044 17.913 17.882 16:001 + 11:047i 19.4 56.8 16.7

Copyright ? 2002 John Wiley & Sons, Ltd.

Second transverse= 26.966 25.340 25.096 25:778 + 7:570i 26.9 28.2

about longitudinal (25.087) (26:830 + 8:010i) (26.9) (28.2)

Note: A: Undamped original 3D FEM model.

Earthquake Engng Struct. Dyn. 2002; 31:19671991

Time history analysis

Recorded - Ch. 7 Computed (case 1) Computed (case 2) Computed (case 3)

_0.5 _0.5 =27.2% _0.5 =17.2% _0.5 =8.1%

0.5 0.5 0.5 0.5

=16.4% =61.1% =26.3%

0.05 0.05 0.05 0.05

_0.05 _0.05 _0.05 _0.05

Figure 3. Records of channel 7 and predictions of Meloland Road Overcrossing response

Recorded - Ch. 8 Computed (case 1) Computed (case 2) Computed (case 3)

=50.5% =4.7% =11.6%

0.5 0.5 0.5 0.5

=56.8% =71.1% =26.7%

0.1 0.1 0.1 0.1

0.05 0.05 0.05 0.05

0.05 0.05 0.05 0.05

Figure 4. Records of channel 8 and predictions of Meloland Road Overcrossing response

RESPONSE OF THE PAINTER STREET OVERCROSSING

Recorded - Ch. 13 Computed (case 1) Computed (case 2) Computed (case 3)

=10.2% =30.1% =18.5%

0.5 0.5 0.5 0.5

=25.4% =75.9% =26.8%

0.05 0.05 0.05 0.05

0.05 0.05 =15.1%

Figure 5. Records of channel 13 and predictions of Meloland Road Overcrossing response

Monolithic embankments and viscoelastic Viscoelastic embankments and monolithic

recorded crest motions recorded crest motions

Figure 6. Bridge models with dierent support idealizations.

Recorded - Ch. 7 Computed (case 1) Computed (case 2) Computed (case 3)

0.5 0.5 =16.8% 0.5 =1.29% 0.5 =9.92%

0.5 0.5 0.5 0.5

=5.64% =39.1% =25.7%

0.05 0.05 0.05 0.05

0.05 0.05 =4.63%

Figure 7. Records of channel 7 and predictions of Meloland Road Overcrossing response

Time history analysis

f1 = 1.78 Hz ( 2.69 Hz)

f2 = 2.29 Hz ( 3.30 Hz)

f3 = 2.60 Hz ( 4.00 Hz) f3 = 2.93 Hz ( 4.45 Hz )

f5 = 3.29 Hz ( 5.24 Hz)

f6 = 3.43 Hz ( 6.61 Hz)

Zhang Et Al-2002-Earthquake Engineering & Structural Dynamics

Zhang Et Al-2002-Earthquake Engineering & Structural Dynamics